The OSS GAS-Rugged (GAS-R) offers unprecedented compute density, performance, and flexibility in the first 10 PetaOPS† AI system for rugged edge computing environments.

Find a Distributor

OSS GAS-R features the world’s most advanced accelerator, the NVIDIA® A100 Tensor Core GPU, enabling AI on the Fly® customers to consolidate training, inference, and analytics into a unified, deployable AI solution at the edge. With available rack and flange mounting options, adaptable power subsystem and unheard-of performance in a modest 23” depth aluminum enclosure the GAS-R excels in demanding autonomous driving vehicle and airborne applications with true “Datacenter in the Sky” capability.

AI on the Fly applications require computing power that can perform an enormous amount of calculations per second. Increasing the compute density of each server node dramatically reduces the number of servers required, resulting in huge savings in cost, power, and space consumed, all tenets of edge AI deployments. Whether your application needs enhanced AI model compute power, high-dimension matrix multiplication, loading and transforming large datasets or scale out inference, eight A100 GPUs with 320 GB high-bandwidth memory, interconnected with six NVSwitches using 3rd generation NVLink at 600GB/s aggregate bandwidth will handle just about any task.

The complete OSS GAS-R solution incorporates building blocks across hardware, networking, software, libraries, and optimized AI models and applications from NGC™. Representing the most powerful end-to-end AI and HPC platform for high-end edge applications, it allows researchers to deliver real-world results in the real world instead of in a sterile environment and deploy solutions into production at scale.


  • 8x NVIDIA® A100 SXM4 GPUs with 4.8TB/s Aggregate NVIDIA® NVLink Bandwidth
  • Up to 64 Core AMD Epyc 7002 or Dual Intel® Scalable Processors
  • 8U x 23” Depth Enclosure
  • Up to 4TB System Memory
  • Over 200TB PCIe Gen4 NVMe™ Storage
  • Four 200Gb/s NICs
  • System Monitoring and Control through BMC
  • MIL-STD 810G tested
  • AC (50-400Hz) and DC Power Options

† INT8 inference with sparsity



14.5”H x 17”W (19” rackmount) x 24.5”D

System Weight


Form Factor

8U rack mount, front rack ears, rear dagger pins

System Storage Capacity

200TB PCIe 4.0 NVMe flash


Up to 64 Core AMD Epyc™ 7002 or Dual Intel® Scalable Processors

Compute GPUs

8x NVIDIA® A100 SXM4 GPUs with NVLINK 3.0



Network Inputs
  • 4x Ethernet & Infiniband up to 200Gb/s, 4x FibreChannel up to 32Gb
  • Other sensor inputs available using FPGA input modules for sensors, radar, digital, analog I/O

Anodized Aluminum

  • Rear Power Supply Input: 100-250VAC (47-63 or 400Hz) or 200-370VDC
  • 2+1, hot-swap power system up to 1600W
LEDs Front Status LEDs displaying system power and IPMI Status
System Monitoring IPMI BMC system monitoring through rear RJ45
Environmental Temperature
  • Operating: 0°C to 35°C
  • Storage (non-operating): -40°C to 71°C
  • -500 to 10,000ft Altitude with rapid decompression
Environmental Humidity
  • Operating: 5% to 95% non-condensing
  • Storage (non-operating): 5% to 100% condensing, after drying
Shock Transport (non-operating): ±10g, 11msec, half-sine pulse, 3 shocks per axis each direction
  • Operational: 5-2000 Hz, .00004-.015 g2/Hz
  • Non-Operational: 5-2000 HZ, .00015-.06 g2/Hz
  • Transportation in optional transport case: 15-2000 Hz, .01-.3 g2/Hz
Noise 80dBA 1m from unit
Standards Tested to MIL-STD-810G, MIL-STD-461E, MIL-STD-464A, MIL-STD-704E, IEC 61000-4-2, conformal coated